下肢运动模式识别及运动姿态预测算法研究

发布时间：2018-06-24 22:56

本文选题：智能下肢假肢 + 神经网络　；参考：《长安大学》2015年硕士论文

【摘要】：智能下肢假肢的研究目的在于改善和提高残疾人的生活质量,促进我国医疗福利事业的发展以及社会的和谐稳定。目前,国内外都已经出现了智能化和仿生程度很高的智能下肢假肢产品,但是价格普遍偏高,难以普及到广大的残疾人群中去。因此,加大力度探索和研发高性能、低成本的智能下肢假肢产品,对于改善我国残疾人的日常生活有着很大的意义。智能下肢假肢的研究必须以下肢运动模式识别以及下肢运动姿态的准确预测为前提。本文主要是在实现下肢运动信息采集的基础上,探索和研究神经网络算法在智能下肢假肢研究领域的可推广性,实现基于神经网络算法的人体下肢运动模式识别以及下肢运动姿态的准确预测,具体所做的研究工作有以下几点:(1)本文为了更好地实现人体下肢运动模式识别以及运动姿态预测,详细分析了人体下肢运动的状态及特征参数,搭建了人体下肢运动的膝关节角度获取系统,并利用角度均值比的方法,对膝关节角度信号进行了简单的归一化处理,将膝关节角度信号转化为膝关节角度特征值。(2)在多运动状态的模式识别上,本文选用了比较成熟的BP神经网络算法及其两种改进算法还有自组织竞争神经网络,一共四种网络分别建立了多运动状态的模式识别模型,并分别对模型进行训练和仿真,最后比较识别结果,发现自组织竞争神经网络建立的模式识别模型识别准确率更好,速度更快,模型训练更加稳定。(3)引入另外一种神经网络算法RBF神经网络,分别利用基于L_M反传算法的BP神经网络以及RBF神经网络建立人体下肢运动姿态预测模型,实现对于人体下肢运动姿态的预测。比较两种模型的仿真结果,发现RBF神经网络建立的人体下肢运动姿态预测模型预测精确,与实际的运动趋势几乎吻合,适用于人体下肢的运动姿态预测。
[Abstract]:The purpose of intelligent limb prosthesis research is to improve and improve the quality of life of the disabled, promote the development of medical welfare and social harmony and stability in China. At present, intelligent limb prosthesis with intelligent and bionic degree has appeared at home and abroad, but the price is generally high, it is difficult to popularize to the majority of the disabled people. Therefore, strengthening the exploration and development of high performance and low cost intelligent limb prosthesis is of great significance for improving the daily life of the disabled in our country. The research of intelligent limb prosthesis must be based on the recognition of the following limb movement pattern and the accurate pretest of the lower limb movement posture. This article is mainly to realize the lower limb movement. On the basis of information collection, the paper explores and studies the generalization of neural network algorithm in the field of intelligent limb prosthesis research, and realizes the recognition of human body movement pattern recognition based on neural network algorithm and the accurate prediction of lower limb movement posture. The specific research work has the following points: (1) this paper is to better realize the lower limb movement of the human body. Dynamic pattern recognition and motion attitude prediction, the state and characteristic parameters of human body movement are analyzed in detail. A knee joint angle acquisition system of lower limb movement is built. The angle signal of knee joint is simplified and normalized by means of angle mean ratio, and the angle signal of knee joint is transformed into knee joint angle special. (2) in the pattern recognition of multi motion state, this paper selects a mature BP neural network algorithm and its two improved algorithms and self-organizing competitive neural network. A total of four kinds of network model recognition model of multi motion state are established respectively, and the model is trained and simulated respectively. Finally, the recognition results are compared and found from the model. The pattern recognition model established by organization competitive neural network has better recognition accuracy, faster speed and more stable model training. (3) introducing another neural network algorithm RBF neural network, using the BP neural network based on the L_M back propagation algorithm and the RBF neural network to establish the human body movement attitude prediction model of the human body, to realize the human body. The prediction of the motion posture of the lower extremities is compared with the simulation results of the two models. It is found that the prediction model of the motion posture of the lower extremities established by the RBF neural network is accurate and almost coincides with the actual movement trend, which is suitable for the motion posture prediction of the lower limbs of the human body.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R496;TP183

【参考文献】